Add checkpoint metadata
Browse files
models/grpo_mustard_dexpolicy_seed0.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"method": "GRPO + DexPolicy",
|
| 3 |
+
"object": "ycb-006_mustard_bottle",
|
| 4 |
+
"seed": 0,
|
| 5 |
+
"source_steps": 507904,
|
| 6 |
+
"additional_steps": 250000,
|
| 7 |
+
"learning_rate": 0.00005,
|
| 8 |
+
"exploration_std_start": 0.10,
|
| 9 |
+
"exploration_std_end": 0.03,
|
| 10 |
+
"credit_module": "grpo_rtg",
|
| 11 |
+
"curriculum_stage": 2
|
| 12 |
+
}
|